Water in cavities

How hydration affects ligand binding in protein cavities is a subtle business. Not only is it still imperfectly understood, but it seems possible that it might be hard to generalize about how the various enthalpic and entropic effects of dehydration of the cavity and the ligand balance out. Yet such issues could be central to the rational design of drugs. Andrew McCammon at UCSD and his coworkers have used MD simulations to try to bring some order to the problem (R. Baron et al., JACS 132, 12091-12097; 2010 – paper here). They point out that both positive and negative entropy changes have been reported previously for water entering protein cavities, and that the result probably depends both on the chemistry and the geometry of the cavity. The key question, they rightly say, is: does water have a passive or active role in cavity-ligand recogniton? They investigate that question using an idealized cavity-ligand combination with various permutations of surface charges on both, looking to develop ‘thermodynamic profiles’ of the binding events. And indeed, the signs and magnitudes of the enthalpy and entropy changes prove to vary widely for the different cases (their Figure 6 tells the story). One finding is that, coincidentally, the net free energy change is similar both for binding driven electrostatically and by hydrophobic interactions.

Similar issues are explored by Hongtao Yu and Steven Rick at the University of New Orleans, who calculate the entropy, enthalpy and free-energy changes on transferring a water molecule from the bulk to various types of protein cavity large enough to hold only a single water (J. Phys. Chem. B 10.1021/jp104209w – paper here). They look in particular at the effects of having different numbers of H-bond donors and acceptors in the cavity. It is again not easy to generalize, but it seems that the thermodynamic consequences of H-bond formation are greater than those exerted, via entropic effects, by the cavity size.

More on how denaturants work, this time from Ilja K. Voets at the University of Fribourg and colleagues (J. Phys. Chem. B 10.1021/jp103515b – paper here). They use light scattering, SANS, NMR, IR and UV-Vis spectroscopy to study how the size and flexibility of the lysozyme molecule changes in a mixed water/DMSO solvent at varying compositional ratios. They see three regimes, in which the protein is compact, wholly unfolded, and partially unfolded, and deduce that there are three major factors influencing these conformational changes: changes in water’s H-bonded structure induced by DMSO, the propensity of DMSO to act as an H-bond acceptor, and DMSO’s role as a poor solvent for polar groups and the polypeptide backbone but a good one for apolar sidechains.

Meanwhile, Yi Qin Gao at Texas A&M and colleagues look at urea denaturation (H. Wei et al., J. Phys. Chem. B 10.1021/jp103770y – paper here). Here the big debate has been whether urea acts indirectly, via its effect on water’s H-bonded network, or directly by interactions with the protein backbone. This group proposed previously that it’s a bit of both. Now they take that idea further by looking at urea denaturation of the chicken villin headpiece protein and some of its mutants. They consider in particular how urea affects the breaking of backbone hydrogen bonds and the penetration of water into the hydrophobic core. In both respects, the influence of urea seems again to be both direct and indirect. For example, the presence of urea seems to enhance water penetration of the core (just as it has been found previously to enhance the hydration of the interiors of carbon nanotubes), and that this enhancement is correlated with binding of urea to the protein surface. Moreover, as the protein unfolds, the exposed hydrophilic regions are stabilized by binding of urea. It seems, then, that the denaturing action cannot be explained by a ‘physical effect’ so much as by a ‘narrative’ of the dynamical process.

The mechanism(s) of anti-freeze glycoproteins also continue to be debated, and Martina Havenith at Bochum and her colleagues raise the intriguing idea that they owe their protective role to an ability to perturb the collective dynamics of water over long ranges, retarding them in the extended hydration shell in a way that suppresses freezing (S. Ebbinghaus et al., JACS 10.1021/ja1051632 – paper here). This proposal, based on terahertz spectroscopic data for an anti-freeze glycoprotein from an Antarctic fish, is quite different from any other that I know of, and could be quite different from the way anti-freeze proteins work – unlike AFP, AFGPs are flexible and don’t have a well-defined structure, so probably do not operate by binding to the surface of ice crystals.

One of my finer moments in my former life at Nature was getting Reza Ghadiri’s work on cyclic peptide nanotubes published in 1993 in the face of scepticism during the review process. The work was soon vindicated, and these self-assembling tubes showed evidence of being able to mediate transport of dissolved species through lipid membranes. Jianfen Fan at Soochow University in Suzhou and colleagues have now used MD to investigate the mechanism of water diffusion through such pores (J. Liu et al., J. Phys. Chem. B 10.1021/jp1039207 – paper here). Water molecules form a linear chain threading a hexapeptide channel, but the H-bonded structure is increasingly three-dimensional for wider cyclic molecules. Understanding the transport process could be valuable for, say, the potential use of self-assembling channels like this in water purification and desalination.

The dynamics and evolution of such water ‘filaments’ in pores and other biological systems, such as lipid membranes, are studied by Marek Orzechowski and Markus Meuwly at the University of Basel, using MD simulations (J. Phys. Chem. B 10.1021/jp1051003 – paper here). Their medium is, however, not explicitly a biological one: it is a monolayer of alkylsilica chains about 1 nm thick, like those used in some chromatographic columns; and the solvent is a water/acetonitrile mixture. They find that water filaments, containing typically tens of molecules, form intermittently in the alkyl layer, and can persist for around 1 ns before being dispersed by thermal fluctuations.

The transition of lipid membranes from a gel to a liquid-crystal phase may significantly alter the membrane’s interactions with water and aqueous ions, say Tomasz Róg at the Tampere University of Technology in Finland and coworkers (M. Stepniewski et al., J. Phys. Chem. B 10.1021/jp104739a – paper here). Both phases might exist in vivo, although the LC phase is the usual one. The authors’ simulations show that in the gel phase the lipid headgroups are partially dehydrated and sodium ions cannot penetrate to the interfacial region to bind to the carbonyl groups.

A new method for estimating hydration free energies of organic molecules with an implicit solvent is described by Maxim Federov of the MPI for Mathematics in the Sciences in Leipzig and coworkers (E. L. Ratkova et al., J. Phys. Chem. B 10.1021/jp103955r – paper here). It’s a modification of the RISM model of Chandler and Anderson, which provides a reasonably computationally cheap way of accounting for solvent structure along with some empirical parametrization of the solvation of specific chemical functionalities (alkyl, hydroxy, carbonyl etc.), ‘trained on’ and tested with small organics. No indication of whether it might be extended to macromolecules, although the authors intend to try it on bioactive compounds.

But the method for predicting hydration structures outlined by Karl Freed at Chicago and colleagues is explicitly geared towards proteins (J. J. Virtanen et al., Biophys. J. 99, 1611-1619; 2010 – paper here). They use simulations of the hydration of ubiquitin, lysozyme and myoglobin to calculate electron radial distribution functions for the different atom types, and then show that these can be used to generate electron densities – and from them, hydration structures – for other proteins. The electron distributions can also be used to calculate X-ray scattering intensity, and the authors intend to use this in future work to compare their predictions with experiment.

It’s the absence of water in the core of the ‘alpha-solenoid’ protein importin-beta that gives this spring-like molecule its astonishing elasticity, according to Helmut Grubmüller and colleagues at the MPI for Biophysical Chemistry in Göttingen (C. Kappel et al., Biophys. J. 99, 1596-1603; 2010 – paper here). Specifically, their simulations indicate that the hydrophobic core has a molten-globule-like conformation that governs its mechanical properties. As such, the protein occupies a middle ground between fully folded and intrinsically disordered, showing how what one might call ‘secondary unstructure’ determined by hydrophobicity allows fine-tuning of a biologically relevant physical property.

And not really ‘water in biology’ at all, but a neat experiment on water at a hydrophilic (mica) surface is reported by Jim Heath at Caltech and colleagues (K. Xu et al., Science 329, 1188-1191; 2010 – paper here). They deposit graphene sheets on mica, and observe islands on the surface with the AFM that appear to be water monolayers ‘sealed in’ by the graphene. These often have faceted edges, suggesting that they are ice-like even at room temperature, and they may be nucleated at surface defects. At 90 percent humidity a water monolayer appears to cover the entire mica surface, and a second adlayer may grow patchily on top.